[NVIDIA] FA3/FA4 Fix (#11606)

Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>

python/sglang/srt/server_args.py

@@ -1071,6 +1071,16 @@ class ServerArgs:
             self.enable_mixed_chunk = False
             self.disable_radix_cache = True
 
+        if self.attention_backend == "fa4" or self.decode_attention_backend == "fa4":
+            raise ValueError(
+                "FA4 backend is only supported for prefill. Please use `--prefill-attention-backend fa4` instead."
+            )
+        if self.prefill_attention_backend == "fa4":
+            logger.warning(
+                f"FA4 backend only supports page size 128, changing page_size from {self.page_size} to 128."
+            )
+            self.page_size = 128
+
     def _handle_page_size(self):
         if self.page_size is None:
             self.page_size = 1

python/sglang/test/test_utils.py

@@ -129,6 +129,11 @@ def is_in_amd_ci():
     return get_bool_env_var("SGLANG_IS_IN_CI_AMD")
 
 
+def is_blackwell_system():
+    """Return whether it is running on a Blackwell (B200) system."""
+    return get_bool_env_var("IS_BLACKWELL")
+
+
 def _use_cached_default_models(model_repo: str):
     cache_dir = os.getenv("DEFAULT_MODEL_CACHE_DIR")
     if cache_dir and model_repo:
@@ -151,6 +156,9 @@ DEFAULT_URL_FOR_TEST = f"http://127.0.0.1:{DEFAULT_PORT_FOR_SRT_TEST_RUNNER + 10
 if is_in_amd_ci():
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH = 3000
 
+if is_blackwell_system():
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH = 3000
+
 
 def call_generate_lightllm(prompt, temperature, max_tokens, stop=None, url=None):
     assert url is not None

sgl-kernel/CMakeLists.txt

@@ -91,7 +91,7 @@ FetchContent_Populate(repo-flashinfer)
 FetchContent_Declare(
     repo-flash-attention
     GIT_REPOSITORY https://github.com/sgl-project/sgl-attn
-    GIT_TAG        f9af0c2a1d82ab1812e6987e9338363cc2bf0f8d
+    GIT_TAG        ff87110aad048bb8c4e6effea4c563ddae88b0eb
     GIT_SHALLOW    OFF
 )
 FetchContent_Populate(repo-flash-attention)
@@ -100,7 +100,7 @@ FetchContent_Populate(repo-flash-attention)
 FetchContent_Declare(
     repo-flash-attention-origin
     GIT_REPOSITORY https://github.com/Dao-AILab/flash-attention.git
-    GIT_TAG        203b9b3dba39d5d08dffb49c09aa622984dff07d
+    GIT_TAG        04adaf0e9028d4bec7073f69e4dfa3f6d3357189
     GIT_SHALLOW    OFF
 )
 FetchContent_Populate(repo-flash-attention-origin)

sgl-kernel/csrc/flash_extension.cc

@@ -23,40 +23,43 @@ TORCH_LIBRARY_FRAGMENT(sgl_kernel, m) {
    * From flash-attention
    */
   m.def(
-      "fwd(Tensor!  q,"
-      "    Tensor   k,"
-      "    Tensor   v,"
-      "    Tensor?  k_new,"
-      "    Tensor?  v_new,"
-      "    Tensor?  q_v,"
-      "    Tensor!? out,"
-      "    Tensor?  cu_seqlens_q,"
-      "    Tensor?  cu_seqlens_k,"
-      "    Tensor?  cu_seqlens_k_new,"
-      "    Tensor?  seqused_q,"
-      "    Tensor?  seqused_k,"
+      "fwd(Tensor   q,"                 // (b, s_q, h, d) or (total_q, h, d) if there is cu_seqlens_q
+      "    Tensor   k,"                 // (b_k, s_k, h_k, d) or (total_k, h_k, d) or paged
+      "    Tensor   v,"                 // (b_k, s_k, h_k, dv) or (total_k, h_k, dv) or paged
+      "    Tensor?  k_new,"             // (b, s_k_new, h_k, d) or (total_k_new, h_k, d)
+      "    Tensor?  v_new,"             // (b, s_k_new, h_k, dv) or (total_k_new, h_k, dv)
+      "    Tensor?  q_v,"               // (b, s_q, h, dv) or (total_q_new, h, dv)
+      "    Tensor?  out,"               // (b, s_q, h, dv) or (total_q, h, dv)
+      "    Tensor?  cu_seqlens_q,"      // b+1
+      "    Tensor?  cu_seqlens_k,"      // b+1
+      "    Tensor?  cu_seqlens_k_new,"  // b+1
+      "    Tensor?  seqused_q,"         // b
+      "    Tensor?  seqused_k,"         // b
       "    int?     max_seqlen_q,"
-      "    int?     max_seqlen_k,"
-      "    Tensor?  page_table,"
-      "    Tensor?  kv_batch_idx,"
-      "    Tensor?  leftpad_k,"
-      "    Tensor?  rotary_cos,"
-      "    Tensor?  rotary_sin,"
-      "    Tensor?  seqlens_rotary,"
-      "    Tensor?  q_descale,"
-      "    Tensor?  k_descale,"
-      "    Tensor?  v_descale,"
-      "    float    softmax_scale,"
+      "    int?     max_seqlen_k,"    // TODO: check if needed
+      "    Tensor?  page_table,"      // (b_k, max_num_pages_per_seq)
+      "    Tensor?  kv_batch_idx,"    // b
+      "    Tensor?  leftpad_k,"       // b
+      "    Tensor?  rotary_cos,"      // seqlen_ro x (rotary_dim / 2)
+      "    Tensor?  rotary_sin,"      // seqlen_ro x (rotary_dim / 2)
+      "    Tensor?  seqlens_rotary,"  // b
+      "    Tensor?  q_descale,"       // (b, h_k)
+      "    Tensor?  k_descale,"       // (b, h_k)
+      "    Tensor?  v_descale,"       // (b, h_k)
+      "    float?   softmax_scale,"   // now optional
       "    bool     is_causal,"
       "    int      window_size_left,"
       "    int      window_size_right,"
-      "    float    softcap,"
+      "    int      attention_chunk,"  // NEW
+      "    float    softcap,"          // promoted to double in C++; schema float is fine
       "    bool     is_rotary_interleaved,"
-      "    Tensor?  scheduler_metadata,"
+      "    Tensor?  scheduler_metadata,"  // (b + 1)
       "    int      num_splits,"
       "    bool?    pack_gqa,"
       "    int      sm_margin,"
-      "    Tensor?  sinks) -> Tensor[]");
+      "    Tensor?  sinks"
+      ") -> (Tensor, Tensor, Tensor, Tensor)");  // NEW return type: tuple of 4 tensors
+
   m.impl("fwd", torch::kCUDA, make_pytorch_shim(&mha_fwd));
 }
 

sgl-kernel/include/sgl_flash_kernel_ops.h

@@ -42,45 +42,44 @@ limitations under the License.
 /*
  * From flash-attention
  */
-std::vector<at::Tensor> mha_fwd(
-    at::Tensor& q,        // (b, s_q, h, d) or (total_q, h, d) if there is cu_seqlens_q
-    const at::Tensor& k,  // (b_k, s_k, h_k, d) or (total_k, h_k, d) if there is cu_seqlens_k or (num_pages, page_size,
+std::tuple<at::Tensor, at::Tensor, at::Tensor, at::Tensor> mha_fwd(
+    at::Tensor q,  // (b, s_q, h, d) or (total_q, h, d) if there is cu_seqlens_q
+    at::Tensor k,  // (b_k, s_k, h_k, d) or (total_k, h_k, d) if there is cu_seqlens_k or (num_pages, page_size,
                    // h_k, d) if there is page_table.
-    const at::Tensor& v,  // (b_k, s_k, h_k, dv) or (total_k, h_k, dv) if there is cu_seqlens_k or (num_pages,
+    at::Tensor v,  // (b_k, s_k, h_k, dv) or (total_k, h_k, dv) if there is cu_seqlens_k or (num_pages,
                    // page_size, h_k, dv) if there is page_table.
-    std::optional<const at::Tensor>&
-        k_new_,  // (b, s_k_new, h_k, d) or (total_k_new, h_k, d) if there is cu_seqlens_k_new
-    std::optional<const at::Tensor>&
-        v_new_,  // (b, s_k_new, h_k, dv) or (total_k_new, h_k, dv) if there is cu_seqlens_k_new
-    std::optional<const at::Tensor>& q_v_,           // (b, s_q, h, dv) or (total_q_new, h, dv) if there is cu_seqlens_q
-    std::optional<at::Tensor>& out_,                 // (b, s_q, h, dv) or (total_q, h, dv) if there is cu_seqlens_q
-    std::optional<const at::Tensor>& cu_seqlens_q_,  // b+1
-    std::optional<const at::Tensor>& cu_seqlens_k_,  // b+1
-    std::optional<const at::Tensor>& cu_seqlens_k_new_,  // b+1
-    std::optional<const at::Tensor>&
+    std::optional<at::Tensor> k_new_,  // (b, s_k_new, h_k, d) or (total_k_new, h_k, d) if there is cu_seqlens_k_new
+    std::optional<at::Tensor> v_new_,  // (b, s_k_new, h_k, dv) or (total_k_new, h_k, dv) if there is cu_seqlens_k_new
+    std::optional<at::Tensor> q_v_,    // (b, s_q, h, dv) or (total_q_new, h, dv) if there is cu_seqlens_q
+    std::optional<at::Tensor> out_,    // (b, s_q, h, dv) or (total_q, h, dv) if there is cu_seqlens_q
+    std::optional<at::Tensor> cu_seqlens_q_,      // b+1
+    std::optional<at::Tensor> cu_seqlens_k_,      // b+1
+    std::optional<at::Tensor> cu_seqlens_k_new_,  // b+1
+    std::optional<at::Tensor>
         seqused_q_,  // b. If given, only this many elements of each batch element's queries and outputs are used.
-    std::optional<const at::Tensor>&
+    std::optional<at::Tensor>
         seqused_k_,  // b. If given, only this many elements of each batch element's keys are used.
-    std::optional<int> max_seqlen_q_,
+    std::optional<int64_t> max_seqlen_q_,
     // TODO: check if we need max_seqlen_k
-    std::optional<int> max_seqlen_k_,
-    std::optional<const at::Tensor>& page_table_,      // (b_k, max_num_pages_per_seq)
-    std::optional<const at::Tensor>& kv_batch_idx_,    // b. indices to index into the KV cache
-    std::optional<const at::Tensor>& leftpad_k_,       // b
-    std::optional<const at::Tensor>& rotary_cos_,      // seqlen_ro x (rotary_dim / 2)
-    std::optional<const at::Tensor>& rotary_sin_,      // seqlen_ro x (rotary_dim / 2)
-    std::optional<const at::Tensor>& seqlens_rotary_,  // b
-    std::optional<at::Tensor>& q_descale_,             // (b, h_k), not (b, h)
-    std::optional<at::Tensor>& k_descale_,             // (b, h_k)
-    std::optional<at::Tensor>& v_descale_,             // (b, h_k)
-    float const softmax_scale,
+    std::optional<int64_t> max_seqlen_k_,
+    std::optional<at::Tensor> page_table_,      // (b_k, max_num_pages_per_seq)
+    std::optional<at::Tensor> kv_batch_idx_,    // b. indices to index into the KV cache
+    std::optional<at::Tensor> leftpad_k_,       // b
+    std::optional<at::Tensor> rotary_cos_,      // seqlen_ro x (rotary_dim / 2)
+    std::optional<at::Tensor> rotary_sin_,      // seqlen_ro x (rotary_dim / 2)
+    std::optional<at::Tensor> seqlens_rotary_,  // b
+    std::optional<at::Tensor> q_descale_,       // (b, h_k), not (b, h)
+    std::optional<at::Tensor> k_descale_,       // (b, h_k)
+    std::optional<at::Tensor> v_descale_,       // (b, h_k)
+    std::optional<double> softmax_scale_,
     bool is_causal,
-    int window_size_left,
-    int window_size_right,
-    float const softcap,
-    bool const is_rotary_interleaved,  // if true, rotary combines indices 0 & 1, else indices 0 & rotary_dim / 2
-    std::optional<at::Tensor>& scheduler_metadata_,  // (b + 1)
-    int num_splits,
+    int64_t window_size_left,
+    int64_t window_size_right,
+    int64_t attention_chunk,
+    double softcap,
+    bool is_rotary_interleaved,  // if true, rotary combines indices 0 & 1, else indices 0 & rotary_dim / 2
+    std::optional<at::Tensor> scheduler_metadata_,  // (b + 1)
+    int64_t num_splits,
     std::optional<bool> pack_gqa_,
-    int const sm_margin,
-    std::optional<const at::Tensor>& sinks_);
+    int64_t sm_margin,
+    std::optional<const at::Tensor>& sinks_);  // (h)

sgl-kernel/python/sgl_kernel/_fa4_interface.py

-# Adapted from https://github.com/Dao-AILab/flash-attention/blob/203b9b3dba39d5d08dffb49c09aa622984dff07d/flash_attn/cute/interface.py
+# Adapted from https://github.com/Dao-AILab/flash-attention/blob/54d8aa6751fc9d5f0357854079261913d5df1f9d/flash_attn/cute/interface.py
 
 # Copyright (c) 2025, Jay Shah, Ganesh Bikshandi, Ying Zhang, Vijay Thakkar, Pradeep Ramani, Tri Dao.
-# [2025-07-04] Version in Cute-DSL, for Hopper and Blackwell. You'd need to install nvidia-cutlass-dsl==4.1.0.
+# [2025-10-14] Version in Cute-DSL, for Hopper and Blackwell. You'd need to install nvidia-cutlass-dsl==4.2.1.
 
 
 import copy
 import gc
 import logging
 import math
-from typing import Optional, Tuple
+from typing import Callable, Optional, Tuple
 
 logger = logging.getLogger(__name__)
 
@@ -18,6 +18,7 @@ import cutlass
 import cutlass.cute as cute
 import torch
 from cutlass.cute.runtime import from_dlpack
+from flash_attn.cute import utils
 from flash_attn.cute.flash_fwd import FlashAttentionForwardSm90
 from flash_attn.cute.flash_fwd_sm100 import FlashAttentionForwardSm100
 
@@ -26,22 +27,6 @@ def maybe_contiguous(x):
     return x.contiguous() if x is not None and x.stride(-1) != 1 else x
 
 
-def _reason_recompile(compile_key, jit_func):
-    compile_cache = jit_func.compile_cache
-    compile_key_map = jit_func.compile_key_map
-    if not compile_cache:
-        return "not compiled yet"
-    for k, v in compile_cache.items():
-        if k == compile_key:
-            continue
-        if len(k) != len(compile_key):
-            continue
-        for i in range(len(k)):
-            if k[i] != compile_key[i]:
-                return f"diff at '{compile_key_map[i]}': {k[i]} vs {compile_key[i]} "
-    return "unknown reason"
-
-
 torch2cute_dtype_map = {
     torch.float16: cutlass.Float16,
     torch.bfloat16: cutlass.BFloat16,
@@ -72,7 +57,11 @@ def _flash_attn_fwd(
     num_threads: int = 384,
     pack_gqa: Optional[bool] = None,
     _compute_capability: Optional[int] = None,
-    return_softmax_lse: Optional[bool] = False,
+    score_mod: Callable | None = None,
+    return_lse: bool = False,
+    out: Optional[torch.Tensor] = None,
+    lse: Optional[torch.Tensor] = None,
+    buffers: Optional[list[torch.Tensor]] = None,
 ) -> Tuple[torch.Tensor, torch.Tensor]:
     q, k, v = [maybe_contiguous(t) for t in (q, k, v)]
     num_head, head_dim = q.shape[-2:]
@@ -169,6 +158,14 @@ def _flash_attn_fwd(
     q_batch_seqlen_shape = (
         (batch_size, seqlen_q) if cu_seqlens_q is None else (total_q,)
     )
+    lse_shape = (
+        (batch_size, num_head, seqlen_q)
+        if cu_seqlens_q is None
+        else (num_head, total_q)
+    )
+    requires_grad = q.requires_grad or k.requires_grad or v.requires_grad
+
+    if out is None:
         out = torch.empty(
             *q_batch_seqlen_shape,
             num_head,
@@ -176,16 +173,36 @@ def _flash_attn_fwd(
             dtype=out_torch_dtype,
             device=device,
         )
-    lse_shape = (
-        (batch_size, num_head, seqlen_q)
-        if cu_seqlens_q is None
-        else (num_head, total_q)
-    )
+    else:
+        expected_out_shape = (*q_batch_seqlen_shape, num_head, head_dim_v)
+        assert (
+            out.shape == expected_out_shape
+        ), f"out tensor shape {out.shape} does not match expected shape {expected_out_shape}"
+        assert (
+            out.dtype == out_torch_dtype
+        ), f"out tensor dtype {out.dtype} does not match expected dtype {out_torch_dtype}"
+        assert (
+            out.device == device
+        ), f"out tensor device {out.device} does not match input device {device}"
+        assert out.is_cuda, "out tensor must be on CUDA device"
+
+    if lse is None:
         lse = (
             torch.empty(lse_shape, dtype=torch.float32, device=device)
-        if return_softmax_lse
+            if requires_grad or return_lse
             else None
         )
+    elif lse is not None:
+        assert (
+            lse.shape == lse_shape
+        ), f"lse tensor shape {lse.shape} does not match expected shape {lse_shape}"
+        assert (
+            lse.dtype == torch.float32
+        ), f"lse tensor dtype {lse.dtype} does not match expected dtype torch.float32"
+        assert (
+            lse.device == device
+        ), f"lse tensor device {lse.device} does not match input device {device}"
+        assert lse.is_cuda, "lse tensor must be on CUDA device"
 
     dtype = torch2cute_dtype_map[q.dtype]
     q_tensor, k_tensor, v_tensor, o_tensor = [
@@ -242,6 +259,7 @@ def _flash_attn_fwd(
     current_stream = cuda.CUstream(torch.cuda.current_stream().cuda_stream)
 
     if compute_capability == 9:  # TODO: tune block size according to hdim
+        # Perf heuristic from upstream: hdim=128, noncausal, non-local benefits from larger n_block
         if head_dim == head_dim_v == 128 and not causal and not local:
             n_block_size = 192
     if compute_capability == 10:
@@ -253,13 +271,34 @@ def _flash_attn_fwd(
         ):
             pack_gqa = False
 
+    if softcap is not None:
+        assert score_mod is None, "softcap and score_mod cannot be used together"
+        score_mod = utils.create_softcap_scoremod(softcap)
+
+    if score_mod is not None:
+        is_varlen = (
+            cu_seqlens_q is not None
+            or cu_seqlens_k is not None
+            or seqused_q is not None
+            or seqused_k is not None
+        )
+        if is_varlen:
+            raise NotImplementedError(
+                "score_mod with buffers is not yet supported for varlen sequences. This will be fixed in a future PR."
+            )
+
+    cute_buffers = None
+    if buffers is not None:
+        cute_buffers = [from_dlpack(buf) for buf in buffers]
+
     compile_key = (
         dtype,
         head_dim,
         head_dim_v,
         qhead_per_kvhead,
         causal,
-        softcap is not None,
+        utils.hash_callable(score_mod) if score_mod is not None else None,
+        buffers is not None,
         lse is None,
         cu_seqlens_q is None,
         cu_seqlens_k is None,
@@ -276,9 +315,6 @@ def _flash_attn_fwd(
         compute_capability,
     )
     if compile_key not in _flash_attn_fwd.compile_cache:
-        logger.info(
-            f"Compiling FA4 kernel with reason: {_reason_recompile(compile_key, _flash_attn_fwd)}"
-        )
         if compute_capability == 9:
             assert page_table is None, "paged KV not supported on SM 9.0"
             # fa_fwd = FlashAttentionForwardSm80(
@@ -290,12 +326,14 @@ def _flash_attn_fwd(
                 is_causal=causal,
                 is_local=local,
                 pack_gqa=pack_gqa,
-                m_block_size=m_block_size,
-                n_block_size=n_block_size,
+                tile_m=m_block_size,
+                tile_n=n_block_size,
                 # num_stages=1,
                 num_stages=2,
                 num_threads=num_threads,
                 Q_in_regs=False,
+                score_mod=score_mod,
+                has_buffers=buffers is not None,
             )
         elif compute_capability == 10:
             assert page_size in [
@@ -313,12 +351,15 @@ def _flash_attn_fwd(
                 and not local
                 and cu_seqlens_q is None
                 and seqused_q is None,
+                score_mod=score_mod,
+                has_buffers=buffers is not None,
             )
         else:
             raise ValueError(
                 f"Unsupported compute capability: {compute_capability}. Supported: 9.x, 10.x"
             )
         # TODO: check @can_implement
+        # TODO caching for buffers; cute_buffers
         _flash_attn_fwd.compile_cache[compile_key] = cute.compile(
             fa_fwd,
             q_tensor,
@@ -333,10 +374,10 @@ def _flash_attn_fwd(
             seqused_q_tensor,
             seqused_k_tensor,
             page_table_tensor,
-            softcap,
             window_size_left,
             window_size_right,
             learnable_sink_tensor,
+            cute_buffers,
         )
     _flash_attn_fwd.compile_cache[compile_key](
         q_tensor,
@@ -351,46 +392,29 @@ def _flash_attn_fwd(
         seqused_q_tensor,
         seqused_k_tensor,
         page_table_tensor,
-        softcap,
         window_size_left,
         window_size_right,
         learnable_sink_tensor,
+        cute_buffers,
     )
     return out, lse
 
 
 _flash_attn_fwd.compile_cache = {}
-_flash_attn_fwd.compile_key_map = [
-    "dtype",
-    "head_dim",
-    "head_dim_v",
-    "qhead_per_kvhead",
-    "causal",
-    "softcap is not None",
-    "lse is None",
-    "cu_seqlens_q is None",
-    "cu_seqlens_k is None",
-    "seqused_q is None",
-    "seqused_k is None",
-    "page_table is not None",
-    "window_size_left is not None",
-    "window_size_right is not None",
-    "learnable_sink is not None",
-    "m_block_size",
-    "n_block_size",
-    "num_threads",
-    "pack_gqa",
-    "compute_capability",
-]
 
 
 def warmup_flash_attn(f):
     """
     Decorator for flash_attn_varlen_func:
-    - On the first call, run several warmup passes with different flag combinations
-    - Warmups are executed sequentially to minimize peak GPU memory usage
-    - Does not modify user-provided tensors (clones data)
-    - Easy to extend with more compile-key dimensions
+    - On first call, run several warmup passes with different flag combinations:
+        * return_softmax_lse in {False, True}
+        * global noncausal (window_size=(None,None))
+        * causal (window_size=(None,0))
+        * local sliding window (window_size=(64,64))
+        * optionally pack_gqa=True if qheads > kvheads and allowed
+    - No score_mod / softcap (not supported for varlen yet)
+    - Executes sequentially to minimize peak GPU mem
+    - Does not modify user tensors (clones)
     """
     done = False
 
@@ -399,30 +423,78 @@ def warmup_flash_attn(f):
 
         def maybe_clone(x):
             if isinstance(x, torch.Tensor):
-                return x.clone()
+                return x.detach().clone()  # detach to avoid autograd edges
             return copy.deepcopy(x)
 
         return tuple(maybe_clone(a) for a in args), {
             k: maybe_clone(v) for k, v in kwargs.items()
         }
 
+    def _infer_heads(args, kwargs):
+        """Infer q and kv head counts from arguments."""
+        # Expect signature: (q, k, v, cu_seqlens_q, cu_seqlens_k, ...)
+        q = args[0] if len(args) > 0 else kwargs.get("q")
+        k = args[1] if len(args) > 1 else kwargs.get("k")
+        try:
+            qh = int(q.shape[-2])
+            kvh = int(k.shape[-2])
+            return qh, kvh
+        except Exception:
+            return None, None
+
     def _run_warmups(args, kwargs):
         """Run warmup calls sequentially and release memory after each."""
         base_args, base_kwargs = _clone_args(args, kwargs)
 
-        # Warmup combinations for return_softmax_lse and causal
-        combos = [
-            dict(return_softmax_lse=False, causal=False),
-            dict(return_softmax_lse=False, causal=True),
-            dict(return_softmax_lse=True, causal=False),
-            dict(return_softmax_lse=True, causal=True),
+        qh, kvh = _infer_heads(base_args, base_kwargs)
+        can_pack_gqa = (
+            qh is not None and kvh is not None and qh % kvh == 0 and qh // kvh > 1
+        )
+        has_page_table = (
+            "page_table" in base_kwargs and base_kwargs["page_table"] is not None
+        )
+
+        # Window presets covering global, causal, and local
+        window_presets = [
+            (None, None),  # global noncausal
+            (None, 0),  # causal
+            (64, 64),  # local sliding window
         ]
 
+        lse_flags = [False, True]
+
+        # Base combo list
+        combos = []
+        for ws in window_presets:
+            for return_lse_flag in lse_flags:
+                combos.append(dict(window_size=ws, return_softmax_lse=return_lse_flag))
+
+        # Optionally add a pack_gqa=True variant (FA4 may disable it internally for some varlen shapes/SMs)
+        if can_pack_gqa:
+            for ws in window_presets:
+                combos.append(
+                    dict(window_size=ws, return_softmax_lse=False, pack_gqa=True)
+                )
+
+        # If page_table is present, warm one combo with it (page_table in compile key for SM100)
+        if has_page_table:
+            combos.append(dict(window_size=(None, None), return_softmax_lse=False))
+
+        # Run sequentially
         for combo in combos:
             wa, wk = _clone_args(base_args, base_kwargs)
+            # Keep user-provided softcap/score_mod OUT (varlen+score_mod unsupported)
+            wk.pop("score_mod", None)
+            if "softcap" in wk and wk["softcap"]:
+                wk["softcap"] = 0.0
+            # Apply combo
             wk.update(combo)
             with torch.cuda.stream(torch.cuda.current_stream()):
+                try:
                     f(*wa, **wk)
+                except Exception as e:
+                    # Some combos can be invalid for specific head dims / arch. Ignore and continue.
+                    logger.debug("Warmup combo skipped: %s", e)
             del wa, wk
             torch.cuda.empty_cache()
             gc.collect()
@@ -430,7 +502,9 @@ def warmup_flash_attn(f):
     def wrapper(*args, **kwargs):
         nonlocal done
         if not done:
-            logger.info("Running flash_attn_varlen_func warmup passes...")
+            logger.info(
+                "Running FA4 warmup (global/causal/local, LSE on/off, optional GQA pack)..."
+            )
             _run_warmups(args, kwargs)
             done = True
         return f(*args, **kwargs)
@@ -472,7 +546,7 @@ def flash_attn_varlen_func(
         learnable_sink=learnable_sink,
         softcap=softcap,
         pack_gqa=pack_gqa,
-        return_softmax_lse=return_softmax_lse,
+        return_lse=return_softmax_lse,
     )
 
     return (out, lse) if return_softmax_lse else out

sgl-kernel/python/sgl_kernel/flash_attn.py

@@ -45,7 +45,7 @@ def flash_attn_with_kvcache(
     qv=None,
     rotary_cos=None,
     rotary_sin=None,
-    cache_seqlens: Optional[Union[(int, torch.Tensor)]] = None,
+    cache_seqlens: Optional[Union[int, torch.Tensor]] = None,
     cache_batch_idx: Optional[torch.Tensor] = None,
     cache_leftpad: Optional[torch.Tensor] = None,
     page_table: Optional[torch.Tensor] = None,
@@ -59,6 +59,7 @@ def flash_attn_with_kvcache(
     softmax_scale=None,
     causal=False,
     window_size=(-1, -1),  # -1 means infinite context window
+    attention_chunk: Optional[int] = None,
     softcap=0.0,  # 0.0 means deactivated
     rotary_interleaved=True,
     scheduler_metadata=None,
@@ -137,6 +138,7 @@ def flash_attn_with_kvcache(
             Default to 1 / sqrt(headdim).
         causal: bool. Whether to apply causal attention mask (e.g., for auto-regressive modeling).
         window_size: (left, right). If not (-1, -1), implements sliding window local attention.
+        attention_chunk: Optional[int]. If not None, splits the query into chunks of this size to save memory.
         softcap: float. Anything > 0 activates softcapping attention.
         rotary_interleaved: bool. Only applicable if rotary_cos and rotary_sin are passed in.
             If True, rotary embedding will combine dimensions 0 & 1, 2 & 3, etc. If False,
@@ -216,6 +218,7 @@ def flash_attn_with_kvcache(
     ]
     rotary_cos, rotary_sin = [maybe_contiguous(x) for x in (rotary_cos, rotary_sin)]
     rotary_seqlens = maybe_contiguous(rotary_seqlens)
+    attention_chunk = 0 if attention_chunk is None else int(attention_chunk)
 
     out, softmax_lse, *rest = torch.ops.sgl_kernel.fwd.default(
         q,
@@ -245,6 +248,7 @@ def flash_attn_with_kvcache(
         causal,
         window_size[0],
         window_size[1],
+        attention_chunk,
         softcap,
         rotary_interleaved,
         scheduler_metadata,
@@ -263,10 +267,11 @@ def flash_attn_varlen_func(
     v,
     cu_seqlens_q,
     cu_seqlens_k,
-    max_seqlen_q,
-    max_seqlen_k,
+    max_seqlen_q=None,
+    max_seqlen_k=None,
     seqused_q=None,
     seqused_k=None,
+    page_table=None,
     softmax_scale=None,
     causal=False,
     qv=None,
@@ -274,6 +279,7 @@ def flash_attn_varlen_func(
     k_descale=None,
     v_descale=None,
     window_size=(-1, -1),
+    attention_chunk=0,
     softcap=0.0,
     num_splits=1,
     pack_gqa=None,
@@ -293,25 +299,18 @@ def flash_attn_varlen_func(
             q,
             k,
             v,
-            cu_seqlens_q,
-            cu_seqlens_k,
-            # max_seqlen_q,
-            # max_seqlen_k,
+            cu_seqlens_q=cu_seqlens_q,
+            cu_seqlens_k=cu_seqlens_k,
             seqused_q=seqused_q,
             seqused_k=seqused_k,
+            page_table=page_table,
             softmax_scale=softmax_scale,
             causal=causal,
-            # qv=qv,
-            # q_descale=q_descale,
-            # k_descale=k_descale,
-            # v_descale=v_descale,
             window_size=window_size,
             softcap=softcap,
-            # num_splits=num_splits,
             pack_gqa=pack_gqa,
-            # sm_margin=sm_margin,
-            return_softmax_lse=return_softmax_lse,
             learnable_sink=sinks,
+            return_softmax_lse=return_softmax_lse,
         )
 
     if not is_fa3_supported():
@@ -319,10 +318,15 @@ def flash_attn_varlen_func(
             "flash_attn at sgl-kernel is only supported on sm90 and above"
         )
 
+    # FA3 requires max_seqlen_q and max_seqlen_k
+    if max_seqlen_q is None or max_seqlen_k is None:
+        raise ValueError("max_seqlen_q and max_seqlen_k are required for FA3")
+
     if softmax_scale is None:
         softmax_scale = (q.shape[-1] + (qv.shape[-1] if qv is not None else 0)) ** (
             -0.5
         )
+    attention_chunk = 0 if attention_chunk is None else int(attention_chunk)
 
     out, softmax_lse, *rest = torch.ops.sgl_kernel.fwd.default(
         q,
@@ -352,6 +356,7 @@ def flash_attn_varlen_func(
         causal,
         window_size[0],
         window_size[1],
+        attention_chunk,
         softcap,
         is_rotary_interleaved=False,
         scheduler_metadata=None,

sgl-kernel/tests/test_flash_attention_4.py

-# Adapted from https://github.com/Dao-AILab/flash-attention/blob/b31ae1e4cd22cf5f820a2995b74b7cd3bd54355a/tests/cute/test_flash_attn.py
+# Adapted from https://github.com/Dao-AILab/flash-attention/blob/8ecf128f683266735ba68e3c106ff67a2611886e/tests/cute/test_flash_attn.py
 
 # Copyright (c) 2025, Jay Shah, Ganesh Bikshandi, Ying Zhang, Vijay Thakkar, Pradeep Ramani, Tri Dao.
 
@@ -10,12 +10,25 @@ import pytest
 import torch
 import torch.nn.functional as F
 from einops import rearrange, repeat
+
+try:
+    from flash_attn.layers.rotary import apply_rotary_emb
+except ImportError:
+    apply_rotary_emb = None
+
 from sgl_kernel.flash_attn import flash_attn_varlen_func, flash_attn_with_kvcache
-from utils import is_hopper
+from sgl_kernel.testing.rotary_embedding import _apply_rotary_emb as apply_rotary_emb
 
+# from utils import is_hopper  # Not used in this test
+
+# Force sgl_kernel.flash_attn wrappers to use FA4 (Cute-DSL) implementations.
+# The wrappers accept a superset of args; for FA4, extra args are ignored.
 flash_attn_varlen_func = partial(flash_attn_varlen_func, ver=4)
 flash_attn_with_kvcache = partial(flash_attn_with_kvcache, ver=4)
 
+# Skip this test on Hopper machine
+skip_condition = torch.cuda.get_device_capability() < (10, 0)
+
 
 def unpad_input(hidden_states, attention_mask, unused_mask=None):
     """
@@ -88,6 +101,11 @@ def generate_random_padding_mask(
         lengths = torch.randint(
             max_seqlen // 3, max_seqlen + 1, (batch_size, 1), device=device
         )
+    else:
+        # This should never happen due to the assertion above, but for linter
+        lengths = torch.full(
+            (batch_size, 1), max_seqlen, device=device, dtype=torch.int32
+        )
 
     if zero_lengths:
         # Generate zero-lengths every 5 batches and the last batch.
@@ -482,8 +500,7 @@ def attention_ref(
 
 
 @pytest.mark.skipif(
-    is_hopper(),
-    reason="skip on hopper",
+    skip_condition, reason="FA4 Requires compute capability of 10 or above."
 )
 # @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float8_e4m3fn])
 @pytest.mark.parametrize("dtype", [torch.bfloat16])
@@ -497,8 +514,8 @@ def attention_ref(
 @pytest.mark.parametrize("deterministic", [False])
 # @pytest.mark.parametrize("softcap", [0.0, 15.0])
 @pytest.mark.parametrize("softcap", [0.0])
-@pytest.mark.parametrize("local", [False, True])
-# @pytest.mark.parametrize("local", [False])
+# @pytest.mark.parametrize("local", [False, True])
+@pytest.mark.parametrize("local", [False])
 @pytest.mark.parametrize("causal", [False, True])
 # @pytest.mark.parametrize("causal", [False])
 # @pytest.mark.parametrize("add_unused_qkv", [False, True])
@@ -522,11 +539,11 @@ def attention_ref(
         (64, 128),
         (128, 128),
         (256, 256),
-        (113, 203),
-        (128, 217),
-        (113, 211),
-        (108, 256),
-        (256, 512),
+        # (113, 203),
+        # (128, 217),
+        # (113, 211),
+        # (108, 256),
+        # (256, 512),
         (307, 256),
         (640, 128),
         (512, 256),
@@ -658,25 +675,7 @@ def test_flash_attn_varlen_output(
         if causal or local:
             key_padding_mask = query_padding_mask
 
-        (
-            q_unpad,
-            k_unpad,
-            v_unpad,
-            qv_unpad,
-            cu_seqlens_q,
-            cu_seqlens_k,
-            seqused_q,
-            seqused_k,
-            max_seqlen_q,
-            max_seqlen_k,
-            q,
-            k,
-            v,
-            qv,
-            output_pad_fn,
-            dq_pad_fn,
-            dk_pad_fn,
-        ) = generate_qkv(
+        result = generate_qkv(
             q,
             k,
             v,
@@ -687,6 +686,25 @@ def test_flash_attn_varlen_output(
             query_unused_mask=query_unused_mask,
             key_unused_mask=key_unused_mask,
         )
+        (
+            q_unpad,  # 0
+            k_unpad,  # 1
+            v_unpad,  # 2
+            qv_unpad,  # 3
+            cu_seqlens_q,  # 4
+            cu_seqlens_k,  # 5
+            seqused_q,  # 6
+            seqused_k,  # 7
+            max_seqlen_q,  # 8
+            max_seqlen_k,  # 9
+            q,  # 10
+            k,  # 11
+            v,  # 12
+            qv,  # 13
+            output_pad_fn,  # 14
+            dq_pad_fn,  # 15
+            dk_pad_fn,  # 16
+        ) = result
         q_unpad, k_unpad, v_unpad = [
             x.detach().to(dtype).requires_grad_() for x in (q_unpad, k_unpad, v_unpad)
         ]
@@ -746,20 +764,16 @@ def test_flash_attn_varlen_output(
                 v_unpad,
                 cu_seqlens_q=cu_seqlens_q,
                 cu_seqlens_k=cu_seqlens_k,
-                max_seqlen_q=None,
-                max_seqlen_k=None,
-                # seqused_q=seqused_q,
-                # seqused_k=seqused_k,
+                # max_seqlen_q and max_seqlen_k not needed for FA4
+                seqused_q=seqused_q,
+                seqused_k=seqused_k,
                 causal=causal,
-                # qv=qv_unpad,
-                # q_descale=q_descale,
-                # k_descale=k_descale, v_descale=v_descale,
                 window_size=window_size,
-                # attention_chunk=attention_chunk,
-                sinks=learnable_sink,
                 softcap=softcap,
+                sinks=learnable_sink,  # FA4 uses learnable_sink, not sinks
                 pack_gqa=pack_gqa,
                 return_softmax_lse=True,
+                ver=4,  # Use FA4
             )
             out = output_pad_fn(out_unpad)
             if query_unused_mask is not None:
@@ -875,8 +889,7 @@ def test_flash_attn_varlen_output(
 
 
 @pytest.mark.skipif(
-    is_hopper(),
-    reason="skip on hopper",
+    skip_condition, reason="FA4 Requires compute capability of 10 or above."
 )
 # @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float8_e4m3fn])
 @pytest.mark.parametrize("dtype", [torch.bfloat16])
@@ -887,8 +900,8 @@ def test_flash_attn_varlen_output(
 # @pytest.mark.parametrize("has_learnable_sink", [False])
 # @pytest.mark.parametrize("new_kv", [False, True])
 @pytest.mark.parametrize("new_kv", [False])
-@pytest.mark.parametrize("local", [False, True])
-# @pytest.mark.parametrize("local", [False])
+# @pytest.mark.parametrize("local", [False, True])
+@pytest.mark.parametrize("local", [False])
 # @pytest.mark.parametrize("causal", [False, True])
 @pytest.mark.parametrize("causal", [True])
 # @pytest.mark.parametrize("seqlen_new_eq_seqlen_q", [True, False])
@@ -900,8 +913,8 @@ def test_flash_attn_varlen_output(
 # @pytest.mark.parametrize("rotary_fraction", [0.0, 0.5, 1.0])
 @pytest.mark.parametrize("rotary_fraction", [0.0])
 # @pytest.mark.parametrize("page_size", [None] + ([1, 4, 128]))
-@pytest.mark.parametrize("page_size", [None, 128])
-# @pytest.mark.parametrize("page_size", [128])
+# @pytest.mark.parametrize("page_size", [None, 128])
+@pytest.mark.parametrize("page_size", [128])
 # @pytest.mark.parametrize("has_leftpad", [False, True])
 @pytest.mark.parametrize("has_leftpad", [False])
 # @pytest.mark.parametrize("has_batch_idx", [False, True])
@@ -1085,6 +1098,7 @@ def test_flash_attn_kvcache(
                 .to(dtype_ref)
             )
             page_table = None
+            num_blocks = None
         else:
             (
                 k_cache,
@@ -1301,31 +1315,24 @@ def test_flash_attn_kvcache(
                 else:
                     k_cache_paged.copy_(k_cache_saved)
                     v_cache_paged.copy_(v_cache_saved)
-                # out, lse, *rest = flash_attn_with_kvcache(
-                out, lse, *rest = flash_attn_with_kvcache(
+                # For FA4, use flash_attn_varlen_func directly instead of flash_attn_with_kvcache
+                # This matches the pattern from the original FA4 test
+                out, lse = flash_attn_varlen_func(
                     q if not varlen_q else q_unpad,
                     k_cache if page_size is None else k_cache_paged,
                     v_cache if page_size is None else v_cache_paged,
-                    # k if not new_kv or not varlen_q else k_unpad,
-                    # v if not new_kv or not varlen_q else v_unpad,
-                    # qv=qv if not varlen_q else qv_unpad,
-                    # rotary_cos=cos,
-                    # rotary_sin=sin,
-                    cache_seqlens=cache_seqlens,
-                    # cache_batch_idx=cache_batch_idx,
-                    # cache_leftpad=cache_leftpad,
-                    page_table=page_table,
                     cu_seqlens_q=cu_seqlens_q,
-                    # cu_seqlens_k_new=cu_seqlens_k_new,
-                    # rotary_seqlens=rotary_seqlens,
+                    cu_seqlens_k=None,  # FA4 doesn't use cu_seqlens_k for KV cache
+                    # max_seqlen_q and max_seqlen_k not needed for FA4
+                    seqused_k=cache_seqlens,  # Use cache_seqlens as seqused_k
+                    page_table=page_table,
                     causal=causal,
                     window_size=window_size,
-                    sinks=learnable_sink,
-                    # attention_chunk=attention_chunk,
-                    # rotary_interleaved=rotary_interleaved,
-                    # scheduler_metadata=scheduler_metadata,
-                    # num_splits=num_splits,
+                    sinks=learnable_sink,  # FA4 uses learnable_sink, not sinks
+                    softcap=0.0,
+                    pack_gqa=None,
                     return_softmax_lse=True,
+                    ver=4,  # Use FA4
                 )
                 if varlen_q:
                     out = output_pad_fn(out)

test/srt/run_suite.py

@@ -169,6 +169,7 @@ suites = {
         TestFile("test_disaggregation_pp.py", 140),
     ],
     "per-commit-4-gpu-b200": [
+        # TestFile("test_flash_attention_4.py"),
         # TestFile("test_gpt_oss_4gpu.py", 600),
         # TestFile("test_deepseek_v3_fp4_4gpu.py", 3600),
     ],

test/srt/test_flash_attention_4.py

+import unittest
+from types import SimpleNamespace
+
+from sglang.srt.environ import envs
+from sglang.srt.utils import get_device_sm, kill_process_tree
+from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.test_utils import (
+    DEFAULT_MODEL_NAME_FOR_TEST,
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    popen_launch_server,
+)
+
+
+@unittest.skipIf(get_device_sm() < 100, "Test requires CUDA SM 100 or higher")
+class TestFlashAttention4(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = DEFAULT_MODEL_NAME_FOR_TEST
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--trust-remote-code",
+            "--mem-fraction-static",
+            "0.8",
+            "--prefill-attention-backend",
+            "fa4",
+        ]
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=other_args,
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_gsm8k(self):
+        args = SimpleNamespace(
+            num_shots=4,
+            data_path=None,
+            num_questions=100,
+            max_new_tokens=512,
+            parallel=128,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval_few_shot_gsm8k(args)
+        print(metrics)
+
+        self.assertGreater(metrics["accuracy"], 0.65)
+
+
+if __name__ == "__main__":
+    unittest.main()